The only instance of social contract I know is Debian's, initially from 1997.

https://www.debian.org/social_contract

>I wonder if this always used to be the case

As written in the article of discussion, it used to be, well, quite a mess. There wasn't an established social expectation that you can ask author to do something, and they will do that. The whole software ecosystem was 100x smaller, and most of the users were tech-savvy. The author released the software somehow, this v1.0 got updated my "many" people (back than many meant 3-4-5), and then, after quite a while, it made a roundtrip back to the author, for which they "officially" released v1.1.

That's it, more or less. If no more bugs found, the software was considered as finished.